aaron/ispc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Add reduce_add() for int8 and int16 types.

Browse Source 
This maps to specialized instructions (e.g. PSADBW) when available.
This commit is contained in:
Matt Pharr
2013-07-25 09:11:39 -07:00
parent 2d063925a1
commit b6df447b55
24 changed files with 464 additions and 44 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
stdlib.ispc
View File

@@ -887,13 +887,32 @@ static inline uniform double select(uniform bool c, uniform double a,
///////////////////////////////////////////////////////////////////////////
// Horizontal ops / reductions
__declspec(safe)
static inline uniform int16 reduce_add(int8 x) {
return __reduce_add_int8(__mask ? x : (int8)0);
}
__declspec(safe)
static inline uniform unsigned int16 reduce_add(unsigned int8 x) {
return __reduce_add_int8(__mask ? x : (int8)0);
}
__declspec(safe)
static inline uniform int32 reduce_add(int16 x) {
return __reduce_add_int16(__mask ? x : (int16)0);
}
__declspec(safe)
static inline uniform unsigned int32 reduce_add(unsigned int16 x) {
return __reduce_add_int16(__mask ? x : (int16)0);
}
__declspec(safe)
static inline uniform float reduce_add(float x) {
// zero the lanes where the mask is off
return __reduce_add_float(__mask ? x : 0.);
}
__declspec(safe)
static inline uniform float reduce_min(float v) {
// For the lanes where the mask is off, replace the given value with
@@ -915,7 +934,7 @@ static inline uniform float reduce_max(float v) {
}
__declspec(safe)
static inline uniform int reduce_add(int x) {
static inline uniform int64 reduce_add(int32 x) {
// Zero out the values for lanes that aren't running
return __reduce_add_int32(__mask ? x : 0);
}
@@ -937,7 +956,7 @@ static inline uniform int reduce_max(int v) {
}
__declspec(safe)
static inline uniform unsigned int reduce_add(unsigned int x) {
static inline uniform unsigned int64 reduce_add(unsigned int32 x) {
// Set values for non-running lanes to zero so they don't affect the
// result.
return __reduce_add_int32(__mask ? x : 0);